Well, here's the problem: every part of the circuit is highly interactive.
The input impedance of the mixer should be around 5K at a minimum in order to work with a wide variety of equipment. Lower than that is no problem for most true professional equipment, but most definitely a problem for consumer and "prosumer" stuff. And the output of the mixer needs to have a level and source impedance compatible with the input of a mic preamp.
The input impedance of the panpot circuit is fairly constant at about 50% of the value of the pot--and this is much better than the "usual" dual-lin panpot circuit, which has an input Z that's all over the place. The panpot needs to be fed from a source impedance of not more than 20% of the pot value. That, along with the minimum input impedance requirement, sets the limit on the value of the level pot.
The panpot needs to be loaded by a resistance equal to the pot value; this consists of the mixing resistor and the buss impedance in series. A 200-Ohm buss impedance is negligible compared to the value of the mixing resistor in the case of this design, so we can ignore it and use a 10K mixing resistor for a 10K panpot, 20K for a 20K pot and so on.
The impedance of the buss is all channels (mixing resistor in series with the panpot source impedance, which varies from zero to about 15% of the pot value) in parallel, which is forced down to the desired value by Rshunt.
If we scale everything up to allow a 50K dual for the pan--which means ~50K mixing resistors-- the output impedance will be too high. If we adjust the shunt on the output to compensate, the total attenuation will be ridiculous, more than twice the already-substantial figure of the circuit as I drew it. You'd be lucky to achieve anything approaching an acceptable signal-to-noise ratio. If we scale it down to use a 10K panpot, the input impedance will be too low to be acceptable for "general purpose" use.
All of this applies to the usual "summing box" case in 2005, where the goal is to mix a number of line-level sources into a mic preamp. But Svart, I'm not really clear on what you're looking to do. You want to gut an existing console and install a passive mix buss? If that's the case, outside of the question of why you would want to do such a thing, the requirement is different when you're using dedicated, built-in amplifiers. For one thing, you usually wouldn't need to force the output Z and level quite so low--although doing so does yield a benefit in reduced crosstalk, at the cost of worsened S/N.
You have to think about what you're hoping to gain from passive mixing and not just do it for the sake of it being "passive" and therefore somehow "better." Study the problems of passive mixing in detail and it becomes obvious why active mixing took off the way it did, once it became easy to accomplish with gobs of cheap solid-state gain. It solves a few problems, although it introduces some new ones of its own! There's always a price to pay no matter how you mix.
FWIW, 20K seems to be a more common value in the UK than here in the States. You might be able to obtain 20KB duals without too much trouble from a UK supplier. And I believe Omeg (also in the UK) will do small runs of custom types at a reasonable price. A stepped pan is not out of the question, either. You already know that Mouser sells a 2P12T MBB rotary at a very reasonable price, if 11 panning positions are good enough.